Context-Aware Collaborative-Intelligence with Spatio-Temporal In-Sensor-Analytics in a Large-Area IoT Testbed

• DOI: 10.48550/arxiv.2005.13003
• Publication Date: 2020-01-01
• Authors: Baibhab Chatterjee; Dong‐Hyun Seo; Shramana Chakraborty; Shitij Avlani; Xiaofan Jiang; Heng Zhang; Mustafa Abdallah; Nithin Raghunathan; Charilaos Mousoulis; Ali Shakouri; Saurabh Bagchi; Dimitrios Peroulis; Shreyas Sen
• Keywords: Testbed; Business Intelligence; Context awareness

Decades of continuous scaling has reduced the energy unit computing to virtually zero, while energy-efficient communication remained primary bottleneck in achieving fully energy-autonomous IoT nodes. This paper presents and analyzes trade-offs between energies required for computation a wireless sensor network, deployed mesh architecture over 2400-acre university campus, is targeted towards multi-sensor measurement temperature, humidity water nitrate concentration smart agriculture. Several scenarios involving In-Sensor-Analytics (ISA), Collaborative Intelligence (CI) Context-Aware-Switching (CAS) cluster-head during CI been considered. A real-time co-optimization algorithm developed minimizing consumption hence maximizing overall battery lifetime individual Measurement results show that proposed ISA consumes ~467X lower as compared traditional Bluetooth Low Energy (BLE) communication, ~69,500X with Long Range (LoRa) communication. When implemented conjunction LoRa, node increases from mere 4.3 hours 66.6 days 230 mAh coin cell battery, preserving more than 98% total information. The CAS algorithms help extending worst-case by an additional 50%, thereby exhibiting network ~104 days, which >90% theoretical limits posed leakage currents present system, effectively transferring information sampled every second. web-based monitoring system was archive measured data manner, report anomalies data.